* Raul Miller (moth@debian.org) wrote: > I'm disputing this. So far, I offer as evidence the fact that 32 bit > userland has been a crucial element in amd64's success. So far as counter > evidence, I'm getting handwaving and "that's not how I built my machine". Or how a number of other people built theirs... We're showing that there's a number of people out there who currently *use* this solution, and that it's a damn sight better than the non-existent alternatives. As for 32bit userland being crucial for amd64's success- that's very likely to be the case *overall*, but it's *not* necessairly the case for Debian's userbase, which is what we're primairly concerned with. And, likely, there's a big difference between the two. > > Fact of life: multiarch is vapour and will not be usable for quite a while. > > I'm talking about 64/32 bit userland -- which is something other > distributions already offer. > > That's not vapor. It is with Debian though, and will continue to be for quite some time. Talking about other distributions isn't terribly useful. > > Care to explain how not having any 64bit userland would be better? > > It'll be a lot easier to support 64/32 bit userland this way. Uh, nope, wrong... We're going to be moving to multiarch on all archs, so this just isn't accurate. > I've got 64+32 bit userland because my toolchain (binutils+gcc+libc) > was built that way. That doesn't work for Debian though, it's no where near that simple. At one point we did have a biarch toolchain almost entirely built, but that's not really the issue here- it's changing all of the library packages which will be quite a bit of pain. Stpehen
Attachment:
signature.asc
Description: Digital signature